| Author | Message | ||
     KABL |
I would like to open a discussion on developing NIR quantitative analysis using "as is" values compared to actual values. Is there limit to the variation of the other components when using "as is" values that would prevent good quantitative calibration development? What is that limit if there is one? How is the additional expected error calculated? For example if the moisture varies by 2%, but is unaccounted for what is the mathematical effect on the quantitative values of other components? | ||
| hlmark |
KABL - there was a fair amount of work done on that problem in the early days of NIR, in the measurement of wheat. In fact, the concept precedes the NIR measurements, since the "actual value" really changes as moisture is gained or lost, and has nothing to do with the measurement. As far as I know, in wheat marketing today it is still done for the protein value, since nobody wants to buy water at the price of wheat (or anything else). Corrections were made to "dry basis" or to some other specific basis that was agreed on. The formula for correcting dry basis is relatively simple: Corrected prot = Uncorrected prot * 100 / (100 - % moisture) Since the moisture appears in the denominator, the correction is non-linear, although over samll moisture ranges a linear correction may be "close enough" The formula to correct it to some other, constant basis, is similar: just subtract that basis value from the 100 in the numerator. Howard \o/ /_\ |